It seems like the "random" samples for focal blur are now the same on every
frame of an animation. Is there any way to randomize it per frame?

I don't mind a bit of noise but if it's not moving it looks like I'm viewing the
scene through a distorted pane of glass. I'm pretty sure it didn't used to have
this problem (though admittedly I don't often use blur in animations).

Post a reply to this message

Le 10/11/2010 14:42, Tek a écrit :
> It seems like the "random" samples for focal blur are now the same on every
> frame of an animation. Is there any way to randomize it per frame?

Can we have a demo setting to try to reproduce it ?
(with a minimal scene, please)

> 
> I don't mind a bit of noise but if it's not moving it looks like I'm viewing the
> scene through a distorted pane of glass. I'm pretty sure it didn't used to have
> this problem (though admittedly I don't often use blur in animations).


-- 
A good Manager will take you
through the forest, no mater what.
A Leader will take time to climb on a
Tree and say 'This is the wrong forest'.

Post a reply to this message

Here's a simple example. Render a few frames of this (e.g. +kfi0 +kff20)

//---start of code
#default { pigment { rgb 1 } finish { diffuse .8 emission .2 } }
background { rgb z }
plane { y, -1 pigment { checker rgb 1, rgb .8 } }
light_source { <-2,3,-1>, rgb 1 }

#declare T = frame_number/final_frame;

box { -1,1 translate x*(1-2*T) pigment { rgb x } }

camera {
 right x*image_width/image_height
 up y
 direction z*.8

 location <-1.5,.5,-3>+x*T
 look_at x*T

 aperture 0.5
 focal_point 0
 blur_samples 4
}
//---end of code

Notice how the noise is completely static. If I had the option to make the noise
different on every frame I'd get a much more visually pleasing result. It's
worth noting that I'm planning to blend multiple frames to get motion blur, so I
was hoping that blending would improve the focal blur too. Though in any case
static noise draws attention to itself far more than if it was moving.


Le_Forgeron <lef### [at] freefr> wrote:
> Le 10/11/2010 14:42, Tek a écrit :
> > It seems like the "random" samples for focal blur are now the same on every
> > frame of an animation. Is there any way to randomize it per frame?
>
> Can we have a demo setting to try to reproduce it ?
> (with a minimal scene, please)
>
> >
> > I don't mind a bit of noise but if it's not moving it looks like I'm viewing the
> > scene through a distorted pane of glass. I'm pretty sure it didn't used to have
> > this problem (though admittedly I don't often use blur in animations).
>
>
> --
> A good Manager will take you
> through the forest, no mater what.
> A Leader will take time to climb on a
> Tree and say 'This is the wrong forest'.

Post a reply to this message

Le 10/11/2010 17:45, Tek nous fit lire :
> Here's a simple example. Render a few frames of this (e.g. +kfi0 +kff20)

Thanks.
Testing with 3.6 for windows, the reported issue was already there.
(the noise at horizon stays the same).
But that change if you raise the blur_samples to 20.
(In fact, even a value of 5 make a change at the horizon)

Post a reply to this message

For 3.7beta, you might have a look at the bokeh setting to vary the sampling

http://wiki.povray.org/content/Documentation:Reference_Section_3

Using a moving pattern of your choice for the distribution of the
samples grid.

Post a reply to this message

Le_Forgeron <jgr### [at] freefr> wrote:
> For 3.7beta, you might have a look at the bokeh setting to vary the sampling
>
> http://wiki.povray.org/content/Documentation:Reference_Section_3
>
> Using a moving pattern of your choice for the distribution of the
> samples grid.

Good idea, but sometimes it still samples the same points (particularly with
very low numbers of samples).

Post a reply to this message

"Tek" <tek### [at] evilsuperbraincom> wrote:
> Le_Forgeron <jgr### [at] freefr> wrote:
> > For 3.7beta, you might have a look at the bokeh setting to vary the sampling
> >
> > http://wiki.povray.org/content/Documentation:Reference_Section_3
> >
> > Using a moving pattern of your choice for the distribution of the
> > samples grid.
>
> Good idea, but sometimes it still samples the same points (particularly with
> very low numbers of samples).

After some testing I'm getting the impression that the bokeh filter isn't
strictly applied. I'm getting square shapes when using a bokeh which is mostly
black, as if the algorithm gives up after a few tries and just uses the sample
anyway. Indeed, using a completely black bokeh filter doesn't cause any error
and I still have blur.

This is an understandable limitation of the bokeh filter implementation, but it
limits my ability to use it to randomize blur samples.

Anyway, for the time being I don't really need high quality focal blur in an
animation, and if I do I can just use more blur_samples to brute-force it. But
it would be nice if someone could take a look at the code to fix these niggles.

Post a reply to this message

Am 10.11.2010 14:42, schrieb Tek:
> It seems like the "random" samples for focal blur are now the same on every
> frame of an animation. Is there any way to randomize it per frame?
>
> I don't mind a bit of noise but if it's not moving it looks like I'm viewing the
> scene through a distorted pane of glass. I'm pretty sure it didn't used to have
> this problem (though admittedly I don't often use blur in animations).

Using the very same focal blur sample sequence for all frames of an 
animation is by design, not by error; I'm pretty sure both earlier 3.7 
betas and 3.6 did the same (they used a pseudo-RNG instead of a Halton 
sequence, but unless I'm totally mistaken they started the PRNG all over 
again with the very same seed for each frame). There appears to be 
consensus among the POV-Ray animators that static noise should be 
favored over dynamic noise.

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> Using the very same focal blur sample sequence for all frames of an 
> animation is by design, not by error; I'm pretty sure both earlier 3.7 
> betas and 3.6 did the same (they used a pseudo-RNG instead of a Halton 
> sequence, but unless I'm totally mistaken they started the PRNG all over 
> again with the very same seed for each frame). There appears to be 
> consensus among the POV-Ray animators that static noise should be 
> favored over dynamic noise.

  How hard would it be to make it optional?

-- 
                                                          - Warp

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
>
> There appears to be
> consensus among the POV-Ray animators that static noise should be
> favored over dynamic noise.

Is there? Can you point me to some discussion of this topic, since I strongly
disagree.

Noise that remains static in screen space when everything else is moving draws
attention to itself. Conversely, animated noise can look quite realistic
provided you have a high enough number of samples for it to pass for film-grain,
and if you don't have enough samples it's going to look bad either way.

Obviously there's no "right" answer to which is the best error (they're both
wrong!), but it would be really nice to have the ability to set the random seed
for the focal blur so I can chose if it changes or not.

Incidentally, I'm pretty sure Pixar use animated noise... though I've failed to
Google a reference for that.

Post a reply to this message